Data Bootcamp Section 002 Final Project¶
Professor Zweig¶
Jarrod Horan & Yuan Huang¶
Data:¶
https://fred.stlouisfed.org/series/DTWEXAFEGS
https://fred.stlouisfed.org/series/DTWEXEMEGS
https://fred.stlouisfed.org/series/DEXCHUS
https://www.kaggle.com/cristiangarrido/covid19geographicdistributionworldwide?select=Countries_ISO.csv
https://www.kaggle.com/fireballbyedimyrnmom/us-counties-covid-19-dataset
https://www.kaggle.com/aayushmishra1512/faang-complete-stock-data
In [1]:
import pandas as pd
import matplotlib.pyplot as plt
import datetime as dt
import seaborn as sns
from numpy.polynomial.polynomial import polyfit
import matplotlib.gridspec as gridspec
import numpy as np
import matplotlib as mpl
mpl.rcParams.update(mpl.rcParamsDefault)
import math
import os
import requests, io
import zipfile as zf
import shutil
import statsmodels.formula.api as smf
import matplotlib.ticker as mtick
from sklearn.neighbors import KNeighborsRegressor as knn
from sklearn.model_selection import train_test_split
from sklearn.model_selection import cross_val_score
from sklearn.linear_model import LinearRegression as linreg
from scipy.stats import kde
import matplotlib.dates as mdates
from sklearn.preprocessing import MinMaxScaler
import plotly.graph_objects as go
%matplotlib inline
Import the datasets¶
In [2]:
path_world_covid = 'COVID-19-geographic-disbtribution-worldwide.csv'
world_covid = pd.read_csv(path_world_covid)
world_covid['dateRep'] = pd.to_datetime(world_covid['dateRep'])
world_covid
Out[2]:
In [3]:
world_covid['geoId'].nunique()
world_covid['countriesAndTerritories'].nunique()
Out[3]:
In [4]:
#drop the countries that aren't in world COVID
path_country_coordinate = 'countries.csv'
country_coord = pd.read_csv(path_country_coordinate)
#country_coord
In [5]:
path_us_counties = 'us-counties.csv'
us_counties = pd.read_csv(path_us_counties)
us_counties['date'] = pd.to_datetime(us_counties['date'])
#us_counties
In [6]:
path_netflix = 'Netflix.csv'
netflix = pd.read_csv(path_netflix).tail(365)
netflix['Date'] = pd.to_datetime(netflix['Date'])
netflix = netflix[netflix['Date'].dt.year == 2020]
#pd.reset_option('display.max_rows', None)
#pd.set_option('display.max_rows', None)
#netflix
In [7]:
path_amazon = 'Amazon.csv'
amazon = pd.read_csv(path_amazon).tail(365)
amazon['Date'] = pd.to_datetime(amazon['Date'])
amazon = amazon[amazon['Date'].dt.year == 2020]
#pd.reset_option('display.max_rows', None)
#amazon
In [8]:
path_google = 'Google.csv'
google = pd.read_csv(path_google).tail(365)
google['Date'] = pd.to_datetime(google['Date'])
google = google[google['Date'].dt.year == 2020]
#pd.reset_option('display.max_rows', None)
#google
In [9]:
path_apple = 'Apple.csv'
apple = pd.read_csv(path_apple).tail(365)
apple['Date'] = pd.to_datetime(apple['Date'])
apple = apple[apple['Date'].dt.year == 2020]
#pd.reset_option('display.max_rows', None)
#apple
In [10]:
path_facebook = 'Facebook.csv'
facebook = pd.read_csv(path_facebook).tail(365)
facebook = facebook
facebook['Date'] = pd.to_datetime(facebook['Date'])
facebook = facebook[facebook['Date'].dt.year == 2020]
#pd.reset_option('display.max_rows', None)
#facebook
In [11]:
path_usd_adv_econ = 'DTWEXAFEGS.csv'
usd_adv_econ = pd.read_csv(path_usd_adv_econ)
usd_adv_econ['DATE'] = pd.to_datetime(usd_adv_econ['DATE'])
usd_adv_econ = usd_adv_econ[usd_adv_econ['DATE'].dt.year == 2020]
#pd.reset_option('display.max_rows', None)
#usd_adv_econ
In [12]:
path_usd_em_econ = 'DTWEXEMEGS.csv'
usd_em_econ = pd.read_csv(path_usd_em_econ)
usd_em_econ['DATE'] = pd.to_datetime(usd_em_econ['DATE'])
usd_em_econ = usd_em_econ[usd_em_econ['DATE'].dt.year == 2020]
#usd_em_econ
In [13]:
path_usd_rmb = 'DEXCHUS.csv'
usd_rmb = pd.read_csv(path_usd_rmb)
usd_rmb['DATE'] = pd.to_datetime(usd_rmb['DATE'])
usd_rmb = usd_rmb[usd_rmb['DATE'].dt.year == 2020]
#usd_rmb
How to does the FAANGs growth rate react to COVID from April - July?¶
In [14]:
# merge stocks
netflix2 = netflix.drop(columns = ['Open','High','Low','Adj Close','Volume'])
netflix2 = netflix2.rename(columns={"Close": "Close_Netflix"})
google2 = google.drop(columns = ['Open','High','Low','Adj Close','Volume'])
google2 = google2.rename(columns={"Close": "Close_Google"})
amazon2 = amazon.drop(columns = ['Open','High','Low','Adj Close','Volume'])
amazon2 = amazon2.rename(columns={"Close": "Close_Amazon"})
apple2 = apple.drop(columns = ['Open','High','Low','Adj Close','Volume'])
apple2 = apple2.rename(columns={"Close": "Close_Apple"})
facebook2 = facebook.drop(columns = ['Open','High','Low','Adj Close','Volume'])
facebook2 = facebook2.rename(columns={"Close": "Close_FB"})
faangs = facebook2.merge(apple2, on = 'Date', how = 'left')
faangs = faangs.merge(amazon2, on = 'Date', how = 'left')
faangs = faangs.merge(google2, on = 'Date', how = 'left')
faangs = faangs.merge(netflix2, on = 'Date', how = 'left')
faangs['Date'] = faangs['Date'].dt.strftime('%Y-%m-%d')
faangs.drop([147,148,149,150,151,152,153,154,155,156,157,158], inplace = True)
#faangs
In [15]:
faangs['Month'] = faangs.Date.str[6:7]
#April
faangs_april = faangs.loc[faangs['Month'] == '4']
faangs_april['Day_Close'] = (faangs_april['Close_FB'] + faangs_april['Close_Apple'] +
faangs_april['Close_Amazon'] + faangs_april['Close_Google'] +
faangs_april['Close_Netflix'])
faangs_april['Date'] = pd.to_datetime(faangs_april['Date'])
f4_piv = faangs_april.pivot_table(index = 'Date',columns = 'Month',values = 'Day_Close', aggfunc = 'sum')
f4_piv = f4_piv.pct_change()
f4_piv
#May
faangs_may = faangs.loc[faangs['Month'] == '5']
faangs_may['Day_Close'] = (faangs_may['Close_FB'] + faangs_may['Close_Apple'] +
faangs_may['Close_Amazon'] + faangs_may['Close_Google'] +
faangs_may['Close_Netflix'])
faangs_may['Date'] = pd.to_datetime(faangs_may['Date'])
f5_piv = faangs_may.pivot_table(index = 'Date',columns = 'Month',values = 'Day_Close', aggfunc = 'sum')
f5_piv = f5_piv.pct_change()
#f5_piv
#June
faangs_june = faangs.loc[faangs['Month'] == '6']
faangs_june['Day_Close'] = (faangs_june['Close_FB'] + faangs_june['Close_Apple'] +
faangs_june['Close_Amazon'] + faangs_june['Close_Google'] +
faangs_june['Close_Netflix'])
faangs_june['Date'] = pd.to_datetime(faangs_june['Date'])
f6_piv = faangs_june.pivot_table(index = 'Date',columns = 'Month',values = 'Day_Close', aggfunc = 'sum')
f6_piv = f6_piv.pct_change()
#f6_piv
#July
faangs_july = faangs.loc[faangs['Month'] == '7']
faangs_july['Day_Close'] = (faangs_july['Close_FB'] + faangs_july['Close_Apple'] +
faangs_july['Close_Amazon'] + faangs_july['Close_Google'] +
faangs_july['Close_Netflix'])
faangs_july['Date'] = pd.to_datetime(faangs_july['Date'])
f7_piv = faangs_july.pivot_table(index = 'Date',columns = 'Month',values = 'Day_Close', aggfunc = 'sum')
f7_piv = f7_piv.pct_change()
#f7_piv
In [16]:
#f6_piv = f6_piv.reset_index()
#f6_piv.dtypes
#f6_piv.ix[index.to_datetime()]
f6_piv.index
Out[16]:
In [17]:
#fig, ax = plt.subplots()
#ax.axhline(june_mean, color = 'red',linestyle = 'dashed')
#f6_piv.plot.line(ax = ax,x = 'Date', y = '6')
In [18]:
country_coord = country_coord.rename(columns={"country": "geoId"})
#country_coord
In [19]:
covid_location = world_covid.merge(country_coord, on = 'geoId', how = 'left', indicator = True)
covid_location = covid_location[covid_location['_merge']=='both']
covid_location = covid_location.drop(columns = ['day','year',
'popData2019','name','Cumulative_number_for_14_days_of_COVID-19_cases_per_100000'])
covid_location = covid_location.rename(columns={"dateRep": "Date","continentExp":"Continent",
"countriesAndTerritories":"Country","latitude":"Latitude",
"longitude":"Longitude","cases":"New Cases",
'deaths':"Deaths"})
covid_location['month'] = covid_location['month'].astype(str)
covid_location = covid_location.loc[(covid_location['month'].str.contains('4|5|6|7|8') == True),:]
covid_location = covid_location.groupby(['month','Date','Country',
'Continent','Latitude','Longitude','countryterritoryCode'], as_index = False)[['New Cases', 'Deaths']].sum()
covid_location.loc[covid_location['Country'] == 'Afghanistan',:]
covid_location['month'] = covid_location['month'].astype(int)
covid_april = covid_location.loc[covid_location['month'] == 4]
covid_april = covid_april.groupby(['Country','Latitude','Longitude','Continent','month','countryterritoryCode'], as_index = False)[['New Cases', 'Deaths']].sum()
covid_may = covid_location.loc[covid_location['month'] == 5]
covid_may = covid_may.groupby(['Country','Latitude','Longitude','Continent','month','countryterritoryCode'], as_index = False)[['New Cases', 'Deaths']].sum()
covid_june = covid_location.loc[covid_location['month'] == 6]
covid_june = covid_june.groupby(['Country','Latitude','Longitude','Continent','month','countryterritoryCode'], as_index = False)[['New Cases', 'Deaths']].sum()
covid_july = covid_location.loc[covid_location['month'] == 7]
covid_july = covid_july.groupby(['Country','Latitude','Longitude','Continent','month','countryterritoryCode'], as_index = False)[['New Cases', 'Deaths']].sum()
In [20]:
covid_july
Out[20]:
In [21]:
world_covid = world_covid.rename(columns={"dateRep": "Date"})
In [22]:
cmap = mpl.cm.RdYlGn
reversed_cmap = cmap.reversed()
n = mpl.colors.Normalize()
plt.style.use('dark_background')
In [23]:
fig, ax = plt.subplots(ncols = 2,nrows = 4)
plt.subplots_adjust(bottom = .7)
#APRIL
covid_april.plot.scatter(ax = ax[0,0],figsize=(20,30), y='Latitude',x='Longitude',s=covid_april['New Cases'] * .01,
color=reversed_cmap(n(covid_april['Deaths'].values * 2500)),
edgecolors = 'white',alpha = .9)
ax[0,0].set_title("Global COVID-19 (April)", fontsize = 16, fontweight = 'bold')
f4_piv.plot.line(ax = ax[0,1])
ax[0,1].axhline(0, color = 'white')
april_mean = f4_piv['4'].mean()
ax[0,1].axhline(april_mean, color = 'red',linestyle = 'dashed')
ax[0,1].legend(bbox_to_anchor = (1, 1), labels = ['Growth Rate','Zero Growth','Mean Rate'])
ax[0,1].set_title("FAANG Growth Rate (April)", fontsize = 16, fontweight = 'bold')
ax[0,1].set_ylim(-.06, .06)
ax[0,1].set_xlabel(xlabel = '')
ax[0,1].yaxis.set_major_formatter(mtick.PercentFormatter(1.0))
#MAY
covid_may.plot.scatter(ax = ax[1,0], y='Latitude',x='Longitude',s=covid_may['New Cases'] * .01,
color=reversed_cmap(n(covid_may['Deaths'].values * 2500)),
edgecolors = 'white',alpha = .9)
ax[1,0].set_title("Global COVID-19 (May)", fontsize = 16, fontweight = 'bold')
ax[1,0].set_xlabel(xlabel = '')
f5_piv.plot.line(ax = ax[1,1])
ax[1,1].axhline(0, color = 'white')
may_mean = f5_piv['5'].mean()
ax[1,1].axhline(may_mean, color = 'red',linestyle = 'dashed')
ax[1,1].legend().remove()
ax[1,1].set_title("FAANG Growth Rate (May)", fontsize = 16, fontweight = 'bold')
ax[1,1].set_ylim(-.06, .06)
ax[1,1].set_xlabel(xlabel = '')
ax[1,1].yaxis.set_major_formatter(mtick.PercentFormatter(1.0))
#JUNE
covid_june.plot.scatter(ax = ax[2,0], y='Latitude',x='Longitude',s=covid_june['New Cases'] * .01,
color=reversed_cmap(n(covid_june['Deaths'].values * 2500)),
edgecolors = 'white',alpha = .9)
ax[2,0].set_title("Global COVID-19 (June)", fontsize = 16, fontweight = 'bold')
ax[2,0].set_xlabel(xlabel = '')
f6_piv.plot.line(ax = ax[2,1])
ax[2,1].set_title("FAANG Growth Rate (June)", fontsize = 16, fontweight = 'bold')
ax[2,1].axhline(0, color = 'white')
june_mean = f6_piv['6'].mean()
ax[2,1].axhline(june_mean, color = 'red',linestyle = 'dashed')
ax[2,1].legend().remove()
ax[2,1].set_ylim(-.06, .06)
ax[2,1].set_xlabel(xlabel = '')
ax[2,1].yaxis.set_major_formatter(mtick.PercentFormatter(1.0))
#JULY
covid_july.plot.scatter(ax = ax[3,0], y='Latitude',x='Longitude',s=covid_july['New Cases'] * .01,
color=reversed_cmap(n(covid_july['Deaths'].values * 2500)),
edgecolors = 'white',alpha = .9)
ax[3,0].set_title("Global COVID-19 (July)", fontsize = 16, fontweight = 'bold')
f7_piv.plot.line(ax = ax[3,1])
ax[3,1].axhline(0, color = 'white')
july_mean = f7_piv['7'].mean()
ax[3,1].axhline(july_mean, color = 'red',linestyle = 'dashed')
ax[3,1].legend().remove()
ax[3,1].set_title("FAANG Growth Rate (July)", fontsize = 16, fontweight = 'bold')
ax[3,1].set_ylim(-.06, .06)
ax[3,1].yaxis.set_major_formatter(mtick.PercentFormatter(1.0))
plt.savefig('Covid_stocks.png')
Graphs are interactive charts visualizing new cases per country, similar to the scatter plots above.¶
In [24]:
fig_a = go.Figure(data=go.Choropleth(
locations = covid_april['countryterritoryCode'],
z = covid_april['New Cases'],
text = covid_april['Country'],
colorscale = 'RdYlGn',
autocolorscale=False,
reversescale=True,
marker_line_color='darkgray',
marker_line_width=0.5,
colorbar_title = 'New Cases',))
fig_a.update_layout(
title_text='April COVID Cases',
geo=dict(
showframe=False,
showcoastlines=True,
projection_type='equirectangular'
),
annotations = [dict(
x=0.55,
y=0.1,
xref='paper',
yref='paper',
text='Source: Kaggle',
showarrow = False
)]
)
In [25]:
fig_m = go.Figure(data=go.Choropleth(
locations = covid_may['countryterritoryCode'],
z = covid_may['New Cases'],
text = covid_may['Country'],
colorscale = 'RdYlGn',
autocolorscale=False,
reversescale=True,
marker_line_color='darkgray',
marker_line_width=0.5,
colorbar_title = 'New Cases',))
fig_m.update_layout(
title_text='May COVID Cases',
geo=dict(
showframe=False,
showcoastlines=True,
projection_type='equirectangular'
),
annotations = [dict(
x=0.55,
y=0.1,
xref='paper',
yref='paper',
text='Source: Kaggle',
showarrow = False
)]
)
In [26]:
fig_june = go.Figure(data=go.Choropleth(
locations = covid_june['countryterritoryCode'],
z = covid_june['New Cases'],
text = covid_june['Country'],
colorscale = 'RdYlGn',
autocolorscale=False,
reversescale=True,
marker_line_color='darkgray',
marker_line_width=0.5,
colorbar_title = 'New Cases',))
fig_june.update_layout(
title_text='June COVID Cases',
geo=dict(
showframe=False,
showcoastlines=True,
projection_type='equirectangular'
),
annotations = [dict(
x=0.55,
y=0.1,
xref='paper',
yref='paper',
text='Source: Kaggle',
showarrow = False
)]
)
In [27]:
fig_july = go.Figure(data=go.Choropleth(
locations = covid_july['countryterritoryCode'],
z = covid_july['New Cases'],
text = covid_july['Country'],
colorscale = 'RdYlGn',
autocolorscale=False,
reversescale=True,
marker_line_color='darkgray',
marker_line_width=0.5,
colorbar_title = 'New Cases',
))
fig_july.update_layout(
title_text='July COVID Cases',
geo=dict(
showframe=False,
showcoastlines=True,
projection_type='equirectangular'
),
annotations = [dict(
x=0.55,
y=0.1,
xref='paper',
yref='paper',
text='Source: Kaggle',
showarrow = False
)]
)
Analysis¶
The column one scatter plots visualize the relationship between the number of new COVID-19 cases (size of bubble) and the number of COVID-19 related deaths (color of bubble) per country from April to July. Smaller and green bubbles are indicative of fewer new cases and deaths, whereas larger and red bubbles are the opposite. The general trends by continent are as follows:
-In North America the United States experiences a decrease in monthly deaths and cases from April to June. However, in July, the resurgence of new cases in America reaches new heights. Canada sees a general of decreasing cases with consistently low fatalities. In Mexico the graphs indicate an increase in both cases and deaths.
-In South America, Brazil is the serious case of the trend, which is increasing new cases and deaths. Brazil’s situation can be considered a result of the lack of lockdown measures.
-Europe undergoes a decrease in both new cases and deaths during the summer months. Likely a result of strict lockdown measures utilized during the first wave of COVID. Russia is an exception and experiences higher levels of new cases.
-Africa does see an increase in cases, especially in South Africa. However, there does not appear to be a substantial increase in COVID-19 related deaths.
-Both Australia and New Zealand retain both low case growth and fatalities.
-In Asia there is a contrast between East Asia and India/Middle East. East Asian countries have near miniscule new cases and deaths, with exception to Japan, which experienced case growth in July. However, the Middle East and India are the opposite, experiencing both an increase in new cases and deaths throughout the summer. India is the most notable as it appears to have the largest case and death increase in Asia. There is also an increase in new cases in South East Asia countries: Indonesia and the Philippines.
Column two line charts depict the aggregated growth rate of FAANG stocks in the aforementioned time period. The dashed, red line represents the mean growth rate for that month. In April the FAANG growth rate was larger, slightly higher than 1%, however from May-July that rate was depressed and near zero growth (~.3%)
What is surprising is the FAANG growth rate was higher when the United States’ death count was its worst and started to decrease as deaths decreased, but cases rose. Potentially due to stock speculation.
Given that East Asian and European cases are low/decreasing there does not seem to be much of a visual relationship between these continents and the FAANG growth rate. In fact when the new cases were at the highest level in the United States in July, the FAANG growth rate had slightly increased from .32% in June to .35% in July.
Monthly FAANG growth rates are below:
April FAANG Rate: 1.18%
May FAANG Rate: 0.37%
June FAANG Rate: 0.32%
July FAANG Rate: 0.35%
In [28]:
#print("April FAANG Rate: {:.2%}".format(april_mean))
#print("May FAANG Rate: {:.2%}".format(may_mean))
#print("June FAANG Rate: {:.2%}".format(june_mean))
#print("July FAANG Rate: {:.2%}".format(july_mean))
What interactions can we see between U.S. COVID and the USD relative to AFE and EM economies?¶
In [29]:
#us_counties
In [30]:
#usd_em_econ
In [28]:
usd_afe = usd_adv_econ[usd_adv_econ['DATE'].dt.date.astype(str) >= '2020-01-21']
#usd_afe
In [29]:
usd_em = usd_em_econ[usd_em_econ['DATE'].dt.date.astype(str) >= '2020-01-21']
#usd_em
In [30]:
us_covid = us_counties[us_counties['date'].dt.date.astype(str) <= '2020-11-06']
us_covid = us_covid.drop(columns = ['fips'], axis = 1)
us_covid = us_covid.set_index(['date'])
us_covid = us_covid.rename(columns={"date":"Date","state":"State","cases": "Cases","deaths":"Deaths"})
#us_covid = us_covid.iloc[::,:]
#us_covid = us_covid.groupby(['Date','State'])['Cases','Deaths'].sum()
us_covid = us_covid.groupby([(us_covid.index.month)]).sum()
#piv = us_covid.pivot_table(index = 'date',columns = 'State', values = 'Cases', aggfunc = 'sum')
#piv.pct_change()
In [31]:
#us_covid
In [32]:
#usd_afe
In [33]:
usd_em
usd = usd_em.merge(usd_afe, on = 'DATE', how = 'left')
usd['DTWEXAFEGS'] = pd.to_numeric(usd['DTWEXAFEGS'],errors = 'coerce')
usd['DTWEXEMEGS'] = pd.to_numeric(usd['DTWEXEMEGS'],errors = 'coerce')
numeric = usd.copy()
usd.dtypes
numeric['EM_Rolling'] = numeric.iloc[:,1].rolling(window=5).mean()
numeric['AFE_Rolling'] = numeric.iloc[:,2].rolling(window=5).mean()
#numeric
In [34]:
usd_pct = usd.copy()
usd_pct['DTWEXEMEGS'] = pd.to_numeric(usd_pct['DTWEXEMEGS'],errors = 'coerce').pct_change()
usd_pct['DTWEXAFEGS'] = pd.to_numeric(usd_pct['DTWEXAFEGS'],errors = 'coerce').pct_change()
In [35]:
plt.style.use('dark_background')
fig2 = plt.figure(constrained_layout = True, figsize = (15,10))
gs = fig2.add_gridspec(2,2)
covid = fig2.add_subplot(gs[:,0])
us_covid[['Cases','Deaths']].plot(ax = covid, color = ['cyan','red'])
covid.set_yscale('log')
covid.set_title('Monthly U.S. COVID-19 Cases & Deaths', fontsize = 16, fontweight = 'bold', fontstyle = 'oblique')
covid.set_xlabel('Month', fontsize = 14)
covid.set_ylabel('Cases & Deaths (log-scale)', fontsize = 14)
covid.set_xticklabels(['','February', 'April','June','August','October'], fontsize = 10, rotation = 'horizontal')
covid.grid(color = 'white', linestyle = '-.', linewidth = 1, axis = 'y')
### top right
num = fig2.add_subplot(gs[0,1:])
numeric.plot(ax = num, color = 'tab:green', x = 'DATE', y ='DTWEXEMEGS')
numeric.plot(ax = num, color = 'cornflowerblue', x = 'DATE', y = 'DTWEXAFEGS')
numeric.plot(ax = num, color = 'red', x='DATE', y= 'EM_Rolling')
numeric.plot(ax = num, color = 'gold', x='DATE', y= 'AFE_Rolling')
num.legend(title = 'Regions', labels = ['Emerging Markets', """Advanced Foreign
Economies""", """EM 5-Day Moving
Average""", """AFE 5-Day Moving
Average"""], loc='upper right',bbox_to_anchor=(1.36, 1.02))
num.set_title('USD Trade Weighted Indices', fontsize = 16, fontweight = 'bold', fontstyle = 'oblique')
num.set_xlabel('Month', fontsize = 14)
num.grid(color = 'white', linestyle = '-.', linewidth = .5, axis = 'y')
num.set_ylabel('Index Jan 2006 = 100', fontsize = 10)
### bottom right
pct = fig2.add_subplot(gs[1:,1:])
usd_pct.plot(ax = pct, color = 'tab:green',alpha = .8, x = 'DATE', y ='DTWEXEMEGS')
usd_pct.plot(ax = pct, color = 'cornflowerblue', alpha = .8, x = 'DATE', y = 'DTWEXAFEGS')
pct.set_xlabel('Month', fontsize = 14)
pct.axhline(0, color = 'red')
pct.set_title('USD Trade Weighted Indices Daily % Change', fontsize = 16, fontweight = 'bold', fontstyle = 'oblique')
pct.yaxis.set_major_formatter(mtick.PercentFormatter(1.0))
pct.legend(title = 'Regions', labels = ['Emerging Markets', """Advanced Foreign
Economies"""],loc='upper right',bbox_to_anchor=(1.36, 1.02))
pct.grid(color = 'white', linestyle = '-.', linewidth = .5, axis = 'y')
In [136]:
us_death_level = pd.read_csv('daily_death_us.csv')
us_death_level['datetime'] =pd.to_datetime(us_death_level .datetime)
us_death_level = pd.DataFrame(us_death_level.groupby('datetime')['daily_death'].sum())
us_death_level = us_death_level.reset_index('datetime')
#us_death_level
In [134]:
combine = pd.merge(us_death_level, usd_pct,left_on='datetime', right_on="DATE")
combine = combine[['datetime','daily_death','DTWEXEMEGS','DTWEXAFEGS']]
combine = combine.set_index('datetime')
combine
Out[134]:
In [135]:
## Correlation matrix
rs = np.random.RandomState(0)
df = pd.DataFrame(rs.rand(10, 10))
corr = combine.corr()
corr.style.background_gradient(cmap='coolwarm')
Out[135]:
Analysis¶
The first graph depicts daily U.S. COVID cases from 1/21 - 11/6 on the logarithmic scale. It is visible that from February to March cases and deaths more than doubled, experiencing the greatest growth of any month. From March to April the growth slows however both cases and deaths still nearly double. In the months after the curve begins to flatten and in October begins to decrease.
The graph in the top right corner shows a Trade Weighted USD compared to Advanced Foreign Economies (AFE) and Emerging Markets (EM) for goods & services. There is an initial appreciation of the dollar against both regions before depreciation beginning in late March. This depreciation trend continues through the end of the graphed time period. The dollar is weaker in November than in March.
The lower right corner graph visualizes daily growth rates in both Trade Weighted USD vs. AFE and EM for goods & services. Heading into March volatility begins to increase and growth rates will break +2% and nearly break -2%. The USD had both substantial increases and decreases relative to both regions, appreciating 2% against EMs and depreciating nearly -2% to AFEs. From the end of March onwards volatility decreased, and with exception to a few days, remained between -1% and 1%.
What’s interesting is that in March, when the U.S. COVID cases and deaths growth substantially increases there is both an increase in volatility and an appreciation of the USD. However, when COVID cases begin to flatten circa beginning of April both Trade Weighted Indices also flatten until late-May, early-June before a depreciation trend. Volatility also decreases from the beginning of April, remaining in the -1% to 1% bounds throughout the rest of the time period. However, after April, as U.S. case and death curves continue to flatten, the depreciation trend continues. Thus, the depreciation of the USD is likely not solely due to COVID but relates to other factors in the economy at large.
However, during the period the USD actually appreciated to EMs by 3.1615 from 122.1471 to 125.3086 while the USD depreciated to AFEs by 4.6691 from 109.7980 to 105.1289.
Does the USD or COVID case growth rates best explain the FAANG growth rate?¶
In [36]:
usd_rmb.dtypes
usd_rmb.DEXCHUS = pd.to_numeric(usd_rmb.DEXCHUS, errors='coerce')
usd_rmb = usd_rmb.rename(columns={"DATE": "Date"})
usd_rmb = usd_rmb.set_index('Date')
usd_rmb = usd_rmb.loc['2020-01-21':'2020-07-31']
usd_rmb = usd_rmb.reset_index()
#usd_rmb = usd_rmb.drop(columns = ['RMB_Growth_Rate'], axis = 1)
usd_rmb['RMB_GR'] = usd_rmb['DEXCHUS'].pct_change()
usd_rmb = usd_rmb.iloc[1:]
#usd_rmb
In [37]:
#faangs = faangs.drop(columns = ['total', 'Month'], axis = 1)
faangs1 = faangs
faangs1['Total'] = (faangs['Close_FB'] + faangs['Close_Apple'] + faangs['Close_Amazon'] +
faangs['Close_Google'] + faangs['Close_Netflix'])
faangs1['Date'] = pd.to_datetime(faangs1['Date'])
faangs1 = faangs1.set_index('Date')
faangs1 = faangs1.loc['2020-01-21':'2020-07-31']
faangs1 = faangs1.reset_index()
faangs1 = faangs1.groupby('Date', as_index = False)['Total'].sum()
faangs1['Faang_GR'] = faangs1['Total'].pct_change()
faangs1 = faangs1.iloc[1:]
#faangs1
In [38]:
us_covid2 = us_counties.rename(columns={"date":"Date","state":"State","cases": "Cases","deaths":"Deaths"})
us_covid2 = us_covid2.set_index('Date')
us_covid2 = us_covid2.loc['2020-01-21':'2020-07-31']
us_covid2 = us_covid2.reset_index()
us_covid2 = us_covid2.groupby('Date', as_index = False)['Cases'].sum()
us_covid2['Case_GR'] = us_covid2['Cases'].pct_change()
us_covid2 = us_covid2.iloc[1:]
#us_covid2
In [39]:
covid_faang = us_covid2.merge(faangs1, on = 'Date', how = 'left')
covid_faang = covid_faang.dropna()
covid_faang = usd_rmb.merge(covid_faang, on = 'Date', how = 'left')
covid_faang = covid_faang.drop(columns = ['Total','Cases','DEXCHUS'])
covid_faang = covid_faang.dropna()
covid_faang = covid_faang.drop(2)
#covid_faang
In [40]:
reg = linreg().fit(X = covid_faang[['Case_GR']], y = covid_faang['Faang_GR'])
covid_faang['yhat1'] = reg.predict(covid_faang[['Case_GR']])
In [41]:
reg2 = linreg().fit(X = covid_faang[['RMB_GR']], y = covid_faang['Faang_GR'])
covid_faang['yhat2'] = reg.predict(covid_faang[['RMB_GR']])
In [42]:
plt.style.use('dark_background')
fig, ax = plt.subplots(nrows = 1, ncols = 2, figsize = (14,6))
#COVID, FAANG
covid_faang.plot.scatter(ax = ax[0], x = 'Case_GR',y='Faang_GR')
covid_faang.sort_values('Case_GR').set_index('Case_GR')['yhat1'].plot(ax = ax[0], color = 'gold', lw = 4)
vals = ax[0].get_yticks()
ax[0].set_yticklabels(['{:,.2%}'.format(x) for x in vals])
valsx = ax[0].get_xticks()
ax[0].set_xticklabels(['{:,.2%}'.format(y) for y in valsx])
ax[0].set_title('U.S. COVID Case GR v. FAANG GR', fontsize = 18, fontweight = 'bold')
ax[0].set_ylabel('FAANG Stock Growth Rate', fontsize = 14)
ax[0].set_xlabel('U.S. COVID Case Growth Rate', fontsize = 14)
#RMB, FAANG
covid_faang.plot.scatter(ax = ax[1], x = 'RMB_GR', y = 'Faang_GR')
covid_faang.sort_values('RMB_GR').set_index('RMB_GR')['yhat2'].plot(ax = ax[1], color = 'gold',lw = 4)
vals1 = ax[1].get_yticks()
ax[1].set_yticklabels(['{:,.2%}'.format(x) for x in vals1])
valsx1 = ax[1].get_xticks()
ax[1].set_xticklabels(['{:,.2%}'.format(y) for y in valsx1])
ax[1].set_title('USD/RMB ER v. FAANG GR', fontsize = 18, fontweight = 'bold')
ax[1].set_ylabel('USD v. RMB Growth Rate', fontsize = 14)
ax[1].set_xlabel('U.S. COVID Case Growth Rate', fontsize = 14)
Out[42]:
In [43]:
reg.score(X = covid_faang[['Case_GR']], y = covid_faang['Faang_GR'])
Out[43]:
In [44]:
reg2.score(X = covid_faang[['RMB_GR']], y = covid_faang['Faang_GR'])
Out[44]:
In [114]:
covid_faang = covid_faang.set_index('Date')
covid_faang = covid_faang.iloc[:,0:3]
covid_faang
Out[114]:
In [115]:
## Correlation matrix
rs = np.random.RandomState(0)
df = pd.DataFrame(rs.rand(10, 10))
corr = covid_faang.corr()
corr.style.background_gradient(cmap='coolwarm')
Out[115]:
Analysis¶
The graphs depict the relationships between USD v. RMB and FAANG stock growth rates with U.S. COVID case percent change from 1/21 - 9/31. It is visible that increases in COVID growth rates caused a decrease in FAANG growth rate, while a slight increase in USD v. RMB growth rate. While it is expected that increased COVID cases is correlated worse stock performance, it is surprising that increases in COVID growth had a relationship with USD appreciation.
The analysis that US COVID Case GR and FAANGs are negatively correlated. However, there is a slight positive correlation between the the USD v. RMB GR and FAANG GR.
What trends exist between US COVID deaths and the FAANG stocks prices?¶
In [30]:
df = pd.read_csv('daily_death_us.csv')
df['datetime'] =pd.to_datetime(df.datetime)
df.head()
Out[30]:
In [31]:
df_death_daily = df[['datetime','location_name','daily_death']]
#df_death_daily
In [32]:
df_death_daily['month'] = df_death_daily['datetime'].dt.month
#df_death_daily
In [34]:
df_death_daily1 = pd.DataFrame(df_death_daily.groupby(['month','location_name'])['daily_death'].sum())
df_death_daily1 = df_death_daily1.reset_index('month')
df_death_daily1 = df_death_daily1.reset_index('location_name')
#df_death_daily1
In [45]:
sns.set(style="ticks", context="talk")
sns.set(rc={'figure.figsize':(11.7,8.27)})
plt.style.use("dark_background")
sns.violinplot(x="month", y="daily_death", data=df_death_daily1)
plt.legend(bbox_to_anchor=(1.05, 1), loc=2, borderaxespad=0.)
plt.title('Violin Plot for the deaily death cases for different states in US',fontsize=20)
Out[45]:
In [54]:
plt.style.use("dark_background")
#sns.set(style="ticks", context="talk")
sns.lineplot(data=df_death_daily, x="datetime", y="daily_death")
sns.set(rc={'figure.figsize':(11.7,8.27)})
plt.title('State level daily death cases in US - General form',fontsize=20)
Out[54]:
In [53]:
sns.set(rc={'figure.figsize':(11.7,8.27)})
sns.set(style="ticks", context="talk")
plt.style.use("dark_background")
sns.lineplot(data=df_death_daily, x="datetime", y="daily_death",hue='location_name')
plt.legend(bbox_to_anchor=(1.05, 1), loc=2, borderaxespad=0.)
plt.title('Daily death cases for different states',fontsize=20)
Out[53]:
In [50]:
state_level = pd.DataFrame(us_counties.groupby(['date','state'])['deaths'].sum())
state_level = state_level.reset_index('state')
state_level = state_level.reset_index('date')
state_level = state_level.sort_values(by=['date','state'])
state_level = state_level.set_index('date')
#state_level
In [51]:
state = pd.Series(state_level['state'].unique())
#plt.style.available
In [108]:
#plt.style.use('dark_background')
#fig,ax = plt.subplots()
#for i in state:
# state_level.loc[state_level['state'] == i,:]['deaths'].plot(ax =ax,figsize = (20,10),label=i)
# ax.legend(loc='best')
#ax.legend()
#state_level['deaths'].mean().plot(ax =ax,figsize = (10,5))
#state_level['death'].plot(ax =ax,figsize = (10,5))
#plt.plot(state_level['date'],state_level['deaths'],label=state_level['state'])
#plot(state_level['death'], label=state_level['state'])
In [53]:
#state_level.iloc[:,1]
In [54]:
state_level_pivot = pd.pivot_table(state_level,index=['date'],columns=['state'],
values=['deaths'],fill_value=np.nan)
state_level_pivot = state_level_pivot.dropna()
#state_level_pivot
In [55]:
for i in range(0,55):
state_level_pivot['death_'+str(i)] = state_level_pivot.iloc[:,i].rolling(window=7).mean()
#state_level_pivot
In [56]:
import datetime
df2_2 = state_level.reset_index('date')
df2_2['Date'] = pd.to_datetime(df2_2['date'])
df2_2['Date'] = df2_2['Date'].dt.strftime('%d.%m.%Y')
df2_2['month'] = pd.DatetimeIndex(df2_2['Date']).month
df2_2['day'] = pd.DatetimeIndex(df2_2['Date']).day
df2_2['dayofyear'] = pd.DatetimeIndex(df2_2['Date']).dayofyear
df2_2['weekofyear'] = pd.DatetimeIndex(df2_2['Date']).weekofyear
df2_2['weekday'] = pd.DatetimeIndex(df2_2['Date']).weekday
df2_2['quarter'] = pd.DatetimeIndex(df2_2['Date']).quarter
df2_2['is_month_start'] = pd.DatetimeIndex(df2_2['Date']).is_month_start
df2_2['is_month_end'] = pd.DatetimeIndex(df2_2['Date']).is_month_end
df2_2 = df2_2.drop(['Date'], axis = 1)
df2_2 = df2_2.drop(['date'], axis = 1)
df2_2= pd.get_dummies(df2_2, columns=['month'], drop_first=True, prefix='month')
df2_2 = pd.get_dummies(df2_2, columns=['weekday'], drop_first=True, prefix='wday')
df2_2 = pd.get_dummies(df2_2, columns=['quarter'], drop_first=True, prefix='qrtr')
df2_2= pd.get_dummies(df2_2, columns=['is_month_start'], drop_first=True, prefix='m_start')
df2_2 = pd.get_dummies(df2_2, columns=['is_month_end'], drop_first=True, prefix='m_end')
df2_2= pd.get_dummies(df2_2, columns=['state'], drop_first=True, prefix='state')
#df2_2
#df2_2
In [57]:
from sklearn.model_selection import train_test_split
X = df2_2.drop(columns=["deaths"]).values
y = df2_2.deaths.values
X_train, X_holdout, y_train, y_holdout = train_test_split(X, y, shuffle=False, test_size=0.5, random_state = 0)
X_val, X_test, y_val, y_test = train_test_split(X_holdout, y_holdout, shuffle=False, test_size=0.5, random_state = 0)
In [58]:
from sklearn.linear_model import LinearRegression
from sklearn.metrics import mean_absolute_error
model_1 = LinearRegression()
model_1.fit(X_train,y_train)
y_predicted = model_1.predict(X_val)
MAE_sklearn = mean_absolute_error(y_val, y_predicted)
yy = model_1.predict(X_train)
MAE_sklearn_train = mean_absolute_error(y_train, yy)
In [59]:
MAE_sklearn
Out[59]:
In [60]:
us_level = pd.DataFrame(state_level.groupby('date')['deaths'].sum())
#us_level
In [65]:
#path_netflix = 'Netflix.csv'
netflix = pd.read_csv(path_netflix).tail(365)
netflix['Date'] = pd.to_datetime(netflix['Date'])
netflix = netflix[netflix['Date'].dt.year == 2020]
#pd.reset_option('display.max_rows', None)
#pd.set_option('display.max_rows', None)
netflix = netflix.set_index('Date')
#netflix
In [63]:
netflix_sub = netflix.iloc[:,3:4]
#netflix_sub
In [79]:
#path_amazon = 'Amazon.csv'
amazon = pd.read_csv(path_amazon).tail(365)
amazon['Date'] = pd.to_datetime(amazon['Date'])
amazon = amazon[amazon['Date'].dt.year == 2020]
#pd.reset_option('display.max_rows', None)
amazon = amazon.set_index('Date')
In [80]:
amazon_sub = amazon.iloc[:,3:4]
#amazon_sub
In [81]:
amazon_sub1 = amazon_sub.reset_index('Date')
#amazon_sub1
In [82]:
#path_google = '/Users/jarrodhoran/Downloads/Google.csv'
google = pd.read_csv(path_google).tail(365)
google['Date'] = pd.to_datetime(google['Date'])
google = google[google['Date'].dt.year == 2020]
#pd.reset_option('display.max_rows', None)
google = google.set_index('Date')
#google
In [83]:
google_sub = google.iloc[:,3:4]
#google_sub
In [84]:
google_sub1 = google_sub.reset_index('Date')
#google_sub1
In [85]:
#path_apple = '/Users/jarrodhoran/Downloads/Apple.csv'
apple = pd.read_csv(path_apple).tail(365)
apple['Date'] = pd.to_datetime(apple['Date'])
apple = apple[apple['Date'].dt.year == 2020]
#pd.reset_option('display.max_rows', None)
apple = apple.set_index('Date')
#apple
In [86]:
apple_sub = apple.iloc[:,3:4]
apple_sub1 = apple_sub.reset_index('Date')
#apple_sub1
In [87]:
#path_facebook = '/Users/jarrodhoran/Downloads/Facebook.csv'
facebook = pd.read_csv(path_facebook).tail(365)
facebook = facebook
facebook['Date'] = pd.to_datetime(facebook['Date'])
facebook = facebook[facebook['Date'].dt.year == 2020]
#pd.reset_option('display.max_rows', None)
facebook = facebook.set_index('Date')
#facebook
In [88]:
facebook_sub = facebook.iloc[:,3:4]
facebook_sub1 = facebook_sub.reset_index('Date')
#facebook_sub1
In [89]:
fig, ax1 = plt.subplots()
fig.suptitle('Deaths Cases for US by Date & Trends with the FAANG Stocks Prices by Date',fontsize=25)
us_level['deaths'].plot(ax =ax1,figsize = (20,10),lw=3.5,alpha = 0.5)
#us_level['SMA_3'] = us_level.iloc[:,0].rolling(window=3).mean()
#us_level['SMA_4'] = us_level.iloc[:,0].rolling(window=4).mean()
us_level['SMA_5'] = us_level.iloc[:,0].rolling(window=5).mean()
us_level['CMA_5'] = us_level.iloc[:,0].expanding(min_periods=5).mean()
us_level['EMA'] = us_level.iloc[:,0].ewm(span=40,adjust=False).mean()
#us_level['SMA_3'].plot(ax=ax,lw=4)
#us_level['SMA_4'].plot(ax=ax)
us_level['SMA_5'].plot(ax=ax1,alpha = 0.5,lw=3.5)
us_level['CMA_5'].plot(ax=ax1,alpha = 0.5,lw=3.5)
us_level['EMA'].plot(ax=ax1,alpha = 0.5,lw=3.5)
ax1.legend(fontsize=25)
plt.savefig('COVID19.png')
In [90]:
#amazon_sub1
In [91]:
#us_level
In [92]:
netflix_sub1 = netflix_sub.reset_index('Date')
In [93]:
fig,ax = plt.subplots(5,figsize=(20,25),sharex=True)
from statsmodels.tsa.stattools import adfuller
def test_stationarity(timeseries,i,name):
#Determing rolling statistics
timeseries['rolmean'] = timeseries.iloc[:,1:2].rolling(window=12).mean()
timeseries['rolstd']= timeseries.iloc[:,1:2].rolling(window=12).std()
#Plot rolling statistics:
plt.style.use('dark_background')
timeseries.plot(x='Date',y='Close',color='blue',label='Original'+' - '+str(name).title(),ax=ax[i])
timeseries.plot(x='Date',y='rolmean',color='red', label='12 Day Rolling Mean',ax=ax[i])
timeseries.plot(x='Date',y='rolstd',color='orange', label = '12 Day Rolling Std',ax=ax[i])
ax[i].legend(loc='best')
#ax[i].show(block=False)
#ax[0].title('Rolling Mean & Standard Deviation - Netflix')
fig.suptitle('Change of FAANG Stock Price by Date',fontsize = 30)
test_stationarity(netflix_sub1,0,'netflix')
test_stationarity(google_sub1,1,'google')
test_stationarity(apple_sub1,2,'apple')
test_stationarity(amazon_sub1,3,'amazon')
test_stationarity(facebook_sub1,4,'facebook')
plt.savefig('FAANG1.png')
Analysis¶
Through the graph for the US state level daily death case, we could find that bascially contain as the level of 20 people dead per day after June until November. As for the relationship for the FAANG stock closing price and the US level deaths, we could see that the stock price for all five companies were maintain as a relatively stable level. However, all of the price were decrease around the mid-March which is the time of the COVID-19 outbreaked in US. It is also interesing to think about why the stock price were increase when the death toll increase significantly after July.
Which other FAANG stock best predicts Google?¶
In [94]:
netflix_sub1 = netflix_sub1.rename(columns={"Close": "Close_Netflix",'rolmean':'rolmean_Netflix',
'rolstd':'rolstd_Netflix'})
amazon_sub1 = amazon_sub1.rename(columns={"Close": "Close_Amazon",'rolmean':'rolmean_Amazon',
"rolstd":'rolstd_Amazon'})
apple_sub1 = apple_sub1.rename(columns={"Close": "Close_Apple",
'rolmean':'rolmean_Apple',
'rolstd':'rolstd_Apple'})
google_sub1 = google_sub1.rename(columns={"Close": "Close_Google",
"rolmean":"rolmean_Google",
"rolstd":'rolstd_Google'})
facebook_sub1 = facebook_sub1.rename(columns={"Close": "Close_Facebook",
'rolmean':'rolmean_FB',
'rolstd':'rolstd_FB'})
#facebook_sub1
In [95]:
stonks = apple_sub1.merge(google_sub1, on = 'Date', how = 'left')
stonks = stonks.merge(amazon_sub1, on = 'Date', how = 'left')
stonks = stonks.merge(netflix_sub1, on ='Date', how = 'left')
stonks = stonks.merge(facebook_sub1, on ='Date', how = 'left')
#stonks = stonks.rename(columns = {"Close_x": "Close_Apple"})
In [96]:
stonks = stonks.set_index('Date')
In [97]:
normalized_stonks = pd.DataFrame(index = stonks.index)
In [98]:
from sklearn import preprocessing
x = stonks.values #returns a numpy array
min_max_scaler = preprocessing.MinMaxScaler()
x_scaled = min_max_scaler.fit_transform(x)
normalized_stonks = pd.DataFrame(x_scaled)
In [99]:
normalized_stonks = normalized_stonks.rename(columns = {0:"Close_Apple",1:'rolmean_Apple',2:'rolstd_Apple',
3: "Close_Google",4:"rolmean_Google",5:'rolstd_Google',
6: "Close_Amazon",7:'rolmean_Amazon',8:'rolstd_Amazon',
9: "Close_Netflix",10:'rolmean_Netflix',11:'rolstd_Netflix',
12: "Close_Facebook",13:'rolmean_FB',14:'rolstd_FB'})
In [100]:
normalized_stonks=normalized_stonks.dropna()
In [101]:
#normalized_stonks
In [102]:
reg1 = smf.ols('Close_Google ~ Close_Amazon',normalized_stonks).fit()
normalized_stonks['yhat1'] = reg1.predict()
In [103]:
reg2 = smf.ols('Close_Google ~ Close_Apple',normalized_stonks).fit()
normalized_stonks['yhat2'] = reg2.predict()
In [104]:
reg3 = smf.ols('Close_Google ~ Close_Netflix',normalized_stonks).fit()
normalized_stonks['yhat3'] = reg3.predict()
In [105]:
reg4 = smf.ols('Close_Google ~ Close_Facebook',normalized_stonks).fit()
normalized_stonks['yhat4'] = reg4.predict()
In [106]:
fig,ax = plt.subplots(nrows = 2, ncols = 2, figsize = (16,14))
#top left
normalized_stonks.plot.scatter(ax = ax[0,0], x = 'Close_Amazon', y = 'Close_Google')
normalized_stonks.sort_values('Close_Amazon').set_index('Close_Amazon')['yhat1'].plot(ax = ax[0,0],
color = 'orchid',lw = 4)
ax[0,0].set_xlabel('Amazon Daily Close', fontsize = 14)
ax[0,0].set_ylabel('Google Daily Close', fontsize = 14)
ax[0,0].set_title('Amazon & Google', fontsize = 16, fontweight = 'bold')
#top right
normalized_stonks.plot.scatter(ax = ax[0,1], x = 'Close_Apple', y = 'Close_Google')
normalized_stonks.sort_values('Close_Apple').set_index('Close_Apple')['yhat2'].plot(ax = ax[0,1],
color = 'orchid',lw = 4)
ax[0,1].set_xlabel('Apple Daily Close', fontsize = 14)
ax[0,1].set_ylabel('Google Daily Close', fontsize = 14)
ax[0,1].set_title('Apple & Google', fontsize = 16, fontweight = 'bold')
#bottom left
normalized_stonks.plot.scatter(ax = ax[1,1], x = 'Close_Facebook', y = 'Close_Google')
normalized_stonks.sort_values('Close_Facebook').set_index('Close_Facebook')['yhat4'].plot(ax = ax[1,1],
color = 'orchid',lw = 4)
ax[1,1].set_xlabel('Facebook Daily Close', fontsize = 14)
ax[1,1].set_ylabel('Google Daily Close', fontsize = 14)
ax[1,1].set_title('Facebook & Google', fontsize = 16, fontweight = 'bold')
#bottom right
normalized_stonks.plot.scatter(ax = ax[1,0], x = 'Close_Netflix', y = 'Close_Google')
normalized_stonks.sort_values('Close_Netflix').set_index('Close_Netflix')['yhat3'].plot(ax = ax[1,0],
color = ['orchid']
,lw = 4)
ax[1,0].set_xlabel('Netflix Daily Close', fontsize = 14)
ax[1,0].set_ylabel('Google Daily Close', fontsize = 14)
ax[1,0].set_title('Netflix & Google', fontsize = 16, fontweight = 'bold')
Out[106]:
In [107]:
print(reg4.summary(),reg2.summary(),reg1.summary(),reg3.summary())
Analysis¶
The charts above use a normalized end-of-day price for the FAANG stocks. The analysis reveals that when predicting the stock price of Google, the end-of-day price for Facebook is the best predictor with an R^2 of .848. The remaining order is Apple (R^2: .762), Amazon (R^2: .361), and Netflix (R^2: .301). Thus, it is likely that the stock prices for Facebook and Google experienced the most similar percent change throughout the time period.
There is a low P-value, which is indicative that the results are unlikely to occur randomly.